Prepare Dataset+Workload for Human DBA Experiments

[apavlo]

trafficstars

We need to write a script that prepares the different benchmarks that we plan on using in our experimental analysis to be analyzed by a human. They will look at this information and then create a design by hand. We then compare these designs against the designs generated by our tool.

What we need to do for this task is to decide what the two data items will look like that the human will examine:

1. The first item is the schema for the database. We will recreate this information use the metadata collection that we extract from the workload. It should look similar to the DESCRIBE command output in MySQL.
2. The second item is the sample workload. We should just provide a single query example (from each unique query_hash) and a percentage of how often that query type appears in the sample workload.

The output for these two items should be stored in separate files. I would make a new directory in exps called tools and store your script in there.